Vol. 43, No. 4APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Apr. 1982, p. 877-8830099-2240/82/040877-07$02.00/0

Enzyme Immunoassay for Detection of Salmonellae in FoodstSCOTT A. MINNICH,1t PAUL A. HARTMAN,"* AND RICHARD C. HEIMSCH2

Department of Microbiology, Iowa State University, Ames, Iowa 50011,' and Department of Bacteriologyand Biochemistry, University of Idaho, Moscow, Idaho 838432

Received 13 October 1981/Accepted 28 December 1981

An enzyme immunoassay was developed to detect Salmonella in foods.Indirect test protocols were developed for use with microtitration plates or Gilfordmicrocuvettes. Samples from enrichment cultures were mixed with H-specificimmunoglobulin G and allowed to react; unbound antibody was removed by three5-min centrifugation washes; goat anti-rabbit antibody conjugated to alkalinephosphatase was added and allowed to react; and unbound conjugate wasremoved by centrifugation washing as before. Salmonella-positive samples wereindicated by the production of a chromogenic reaction product after the additionof alkaline phosphatase substrate. The color could be read visually or quantifiedby absorbance. Ninety-eight food samples were examined to compare the enzymeimmunoassay with enrichment serology, immunofluorescence, and the Food andDrug Administration pure culture technique. The enzyme immunoassay wassensitive and specific, and it possessed advantages over methods currently in use.Furthermore, when the enzyme immunoassay was used to screen preenrichmentmedia, the results indicated that it might be decidedly more sensitive than theconventional pure culture technique.

The detection of salmonellae in foods andfoodstuff involves a series of enrichment stepsbecause these pathogens, when present in foods,are generally found in low numbers and are oftensublethally injured. Therefore, detection meth-odologies for salmonellae must be sensitive andallow for the resuscitation and growth initiationof injured cells. Two methods of Salmonelladetection are currently recommended by theFood and Drug Administration (3): (i) a pureculture technique (PCT) involving preenrich-ment, selective enrichment, and selective plat-ing; and (ii) immunofluorescence (IF) after se-lective enrichment. The PCT is an expensiveand time-consuming process that requires 4 or 5days for analysis (3). In contrast, IF is morerapid; it is sensitive, but many false-positives areobtained because polyvalent OH antisera areused and anti-O antibodies cross-react with awide spectrum of related enteric bacteria (16,17). Enrichment serology (ES; 13), a third meth-od used by some laboratories, is more specificthan IF, but ES is more cumbersome and has notmet with widespread acceptance. The list ofadditional techniques for the detection of salmo-nellae is lengthy; however, the acceptance of anew procedure requires that it demonstrate de-

t Journal paper no. J-10419 of the Iowa Agriculture andHome Economics Experiment Station, Ames, Iowa, andResearch Paper no. 81519 of the Idaho Agricultural Experi-ment Station, Moscow, Idaho.

t Present address: Department of Biological Sciences, Pur-due University, West Lafayette, IN 47906.

cided advantages over existing methods. Sensi-tivity, technical expertise required, time untilresults are obtained, cost effectiveness, andamenability to automation are factors thatshould be considered.Enzyme-labeled antibodies were used by Kry-

sinski and Heimsch (5) to detect salmonellae onmembrane filters. This technique was sensitive,permitted multisample analysis, and required aminimum of equipment. The major obstacle toperfecting the technique was the requirement fora polyvalent H-antiserum that was free of cross-reacting 0 antibodies. Minnich and Heimsch(S. A. Minnich, M.S. thesis, University of Ida-ho, Moscow, 1978) improved the Krysinski-Heimsch (5) procedure by purifying immuno-globulin G (IgG) from commercial antisera. IgM,which is the primary immunoglobulin class elic-ited by somatic antigens (6, 7), was removed.The use of purified IgG has been recommendedby others for increased specificity of IF methods(15, 17).The present study was undertaken to further

improve an enzyme immunoassay (EIA) to de-tect salmonellae in foods. An indirect EIA thatwas sensitive and specific was developed; theEIA was compared with the PCT, IF, and ESmethods.

MATERIALS AND METHODS

Stock cultures and media. Most of the bacterialcultures were obtained from collections maintained atthe University of Idaho and Iowa State University.

877

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

878 MINNICH, HARTMAN, AND HEIMSCH

Salmonella serotypes utilized in the pure culture stud-ies were obtained from R. W. Ryder, Centers forDisease Control, Atlanta, Ga.Media were commercial products (Difco Labora-

tories, Detroit, Mich.), except M-broth, which wasprepared as described by Sperber and Deibel (13).Antibody preparations. Spicer-Edwards polyvalent

anti-H IgG (Difco) was purified by using staphylococ-cal protein A (SPA) affinity chromatography. A 40-mgportion of SPA (Sigma Chemical Co., St. Louis, Mo.)was coupled to Sepharose 4B (Pharmacia Fine Chemi-cals, Inc., Piscataway, N.J.) as described by March etal. (9) and modified by Swaminathan et al. (15).Routinely, 4 to 6 ml of pooled Spicer-Edwards antise-rum was added to a column (0.5 by 10 cm) containingSPA-Sepharose 4B. The column was washed withphosphate-buffered saline (PBS; pH 7.4), and theprotein content of the eluate was monitored by absor-bance at 280 nm. When the absorbance readingsdropped below 0.05, 0.1 M glycine-HCl buffer, pH 3.0,was added to the buffer reservoir and bound IgG waseluted. The IgG peak was pooled and dialyzed over-night against PBS to bring the pH back to neutrality.Purity of the antibody preparations was checked byusing sodium dodecyl sulfate-polyacrylamide gel elec-trophoresis. The purified IgG was stored at 4°C in PBScontaining 0.2% NaN3. Before use, the purified IgGwas diluted in polyvinyl pyrrolidone-bovine serumalbumin-PBS (2.0 g of polyvinyl pyrrolidone-40, 0.5 gof bovine serum albumin, and 0.6 ml of Formalin in100 ml of PBS; pH 7.2). The appropriate dilution ofpurified IgG was determined by using the highestdilution that produced positive tube agglutination testson three known Salmonella serotypes. Regenerationof the SPA-Sepharose 4B column was accomplishedby flushing the column thoroughly with PBS, pH 7.2

Calf intestinal alkaline phosphatase (type VII; Sig-ma) was conjugated to goat anti-rabbit immunoglob-ulin (GIBCO Laboratories, Grand Island, N.Y.) byusing the procedure described by Voller et al. (18, 19).A commercial conjugate became available during thecourse of this study, and one preparation (Sigma) washighly satisfactory; it possessed a higher titer and wasmore specific than conjugates that we prepared.E1A. An EIA was developed by modifying the

procedure of Polin and Kennett (12). A 0.1-ml amountof enrichment broth was transferred to 10 ml of M-broth and incubated at 35°C for 4 to 6 h; then 0.1 ml ofthe M-broth culture was added to either a Gilfordcuvette (Gilford Instrument Laboratories, Inc., Ober-lin, Ohio) or a microtitration plate well (Micro-Elisaplate, Dynatech Laboratories, Inc., Alexandria, Va.).Unless specified otherwise, three wells were used persample, two for replicate analyses and the third for acontrol. A 0.1-ml portion of purified Spicer-EdwardsIgG that had been diluted in polyvinyl pyrrolidone-bovine serum albumin-PBS was added to two of thetest sample wells. Diluent without IgG was added tothe control well, which served to monitor washingefficiency and nonspecific binding of conjugate. Thecuvettes or microtitration plates were incubated atroom temperature for 1 h and then were rinsed threetimes with PBS-Tween 20 (0.5% Tween 20 in PBS, pH7.2) to remove unbound IgG. Rinsing was accom-plished by centrifugation at 3,000 rpm for 5 min in anInternational centrifuge equipped with a rotor de-signed for microtitration plates; plastic spacers were

made to adapt the plate holders for the Gilford cuvettepacks. A 0.1-ml amount of Sigma goat anti-rabbit-alkaline phosphatase conjugate diluted 1:1,000 wasadded to each well; the plates were incubated for 1 h atroom temperature, and unbound conjugate was re-moved by washing, as before. Finally, each wellreceived 0.2 ml of alkaline phosphatase substrate. Thesubstrate was prepared on the day of use by adding 5mg (one tablet of Sigma phosphatase substrate 104) to5 ml of diethanolamine buffer (97 ml of diethanola-mine, 0.2 g of NaN3, and 100mg of MgCl2-6H2O in 800ml of distilled water; pH adjusted to pH 9.6 with 1 MHCl and volume made to 1 liter). After incubation at35°C for 30 min, the production of a yellow reactionproduct indicated a positive sample. Although thecolor could be observed visually, in some instances toobtain quantitative results we also read absorbanciesat 405 nm by using a Titertek EIA reader (FlowLaboratories, McLean, Va.) or a Gilford manual EIAreader (Gilford Instrument Laboratories, Inc.).

Detection of salmonellae in foods. For this study, atotal of 98 naturally contaminated food samples wereobtained from private and federal agency laboratoriesand local retail food markets. Four different detectionprocedures were compared. One procedure was theEIA described above. IF assays were conducted byusing commercial polyvalent OH antisera (ClinicalSciences, Inc., Whippany, N.J.), as described byThomason (16). ES was performed as described bySperber and Deibel (13); pooled Spicer-Edwards se-rum diluted to give a final concentration of 1:500 wasused. After postenrichment in M-broth, tube aggluti-nations were performed at 50°C, and the reactionswere scored after 4 h. The conventional PCT wasperformed as described by Poelma and Silliker (11).Samples, 25 or 50 g, were preenriched in Trypticasesoy broth (BBL Microbiology Systems, Cockeysville,Md.) (dried yeast), reconstituted nonfat dry milk(chocolate candy), or lactose broth (all other samples).Tetrathionate and selenite-cystine broths were usedfor selective enrichment, and salmonella-shigella, bril-liant green, and bismuth sulfite agars were used forselective plating. Suspect Salmonella-like colonieswere transferred to tubes of lysine iron and triple sugariron agars. Isolates positive for Salmonella in lysineiron and triple sugar iron agars were further character-ized by using the Mini-Tek identification system forEnterobacteriaceae and were also retested by tubeagglutination with Spicer-Edwards antiserum. Isolatesconforming to the definition of the genus Salmonellawere sent to the National Veterinary Service Labora-tory, Ames, Iowa, for serotyping.

Sensitivity of the EIA procedure. Studies with Sal-monella anatum, S. schwarzengrund, and S. rubislawwere conducted to determine the level of sensitivity ofthe EIA procedure. Serial decimal dilutions of over-night M-broth cultures were made in PBS. A viablecell count was made by plating appropriate dilutions inTrypticase soy agar (BBL). Concomitantly, 0.1-mlamounts of dilutions 10-1 through 10-7 were dis-pensed into microtitration plate wells and subjected toEIA analysis. A control, Escherichia coli B, was usedto determine background levels of absorbance. Resultsof the completed tests were scored by visual observa-tion.A preliminary study also was conducted by using a

fluorogenic substrate (20), 4-methyl umbelliferyl phos-

APPL. ENVIRON. MICROBIOL.

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

EIA FOR FOOD-BORNE SALMONELLAE 879

E0-

0CCY_ Ici

02 4 6 8 1012 14 16 18 20 22 24

FRACTION NO (3m1/ FR ACTION)FIG. 1. Fractionation of Spicer-Edwards antise-

rum on an SPA-Sepharose 4B column. The first peak(fractions 1 to 6) is composed of serum proteins otherthan IgG that were eluted in PBS. Glycine-HCl bufferwas added to the buffer reservoir at point "a." Thesecond peak (fractions 14 to 20) contains purified IgG.O.D., Optical density.

phate (Sigma), at a concentration of 25 ,ug/ml indiethanolamine buffer. An overnight broth culture ofS. anatum was diluted as described earlier, and viablecounts and EIA analyses were performed as describedabove. After 30 min of incubation with substrate, theplates were held under a long-wave UV light to detectfluorescence of the reaction product, 4-methyl umbel-liferone.

RESULTSA typical elution profile of Spicer-Edwards

polyvalent antiserum from SPA-Sepharose 4B isshown in Fig. 1. About 60 mg of purified IgG(fractions 15 through 18, Fig. 1) was obtainedfrom 15 ml of antiserum.A sodium dodecyl sulfate-polyacrylamide gel

procedure was used to compare preparations ofwhole Spicer-Edwards serum, commercial goatanti-rabbit IgG, and Spicer-Edwards serum puri-fied by SPA-Sepharose 4B chromatography.Whereas numerous bands were observed in un-purified Spicer-Edwards antiserum (A, Fig. 2),only two protein bands were present in the goatanti-rabbit IgG (B, Fig. 2) and purified Spicer-Edwards (C, Fig. 2). The two bands in thepurified preparations (B and C, Fig. 2) repre-sented the light and heavy chains of IgG. Thepurified Spicer-Edwards IgG agglutinated allSalmonella serotypes tested. No IgM was de-tectable by Ouchterlony double diffusion againstanti-rabbit IgM. Titers of the purified IgG prepa-rations ranged from 1:40 to 1:100, depending onthe amount of material placed on the SPA-Sepharose 4B column.

FIG. 2. Sodium dodecyl sulfate-polyacrylamidegels of: (A) unpurified Spicer-Edwards antiserum, (B)commercial goat anti-rabbit IgG, and (C) IgG purifiedfrom Spicer-Edwards antiserum by passage through acolumn containing SPA-Sepharose 4B.

Table 1 contains results obtained with purecultures by using a Gilford EIA reader. Dupli-cate samples were run with each culture, and acontrol cuvette was used to monitor the efficien-cy of washing of unbound antibody and enzyme-antibody conjugate. Absorbancies of cuvettesthat contained salmonellae all exceeded 0.9, andabsorbancies of control cuvettes and controlcultures were all <0.5. Table 2 contains theresults of a similar study, except the analyseswere conducted in microtitration plates and aFlow Titertek EIA reader was used. Absorban-cies of wells containing salmonellae all exceeded1.0, and absorbancies of control wells were all<0.1. Representative Salmonella-positive and-negative EIA results, using microcuvettes andmicrotitration plates, are shown in Fig. 3 and 4,respectively.A comparison of the EIA procedure conduct-

VOL. 43, 1982

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

880 MINNICH, HARTMAN, AND HEIMSCH

TABLE 1. Absorbancies obtained with a Gilfordinstrument when pure cultures were analyzed by an

EIAAbsorbance

Organism Sample Sample Control"1 2 Cnrl

Salmonella anatum 0.87 1.58 0.10S. blockley 2+ 2+ 0.30S. bredeney 2+ 2+ 0.42S. cerro 2+ 1.84 0.22S. illinois 2+ 2+ 0.49S. kaapstad 1.92 1.98 0.10S. miami 2+ 2+ 0.21S. paratyphi B 2+ 2+ 0.16S. rubislaw 2+ 2+ 0.41S. sandiego 2+ 2+ 0.09S. senftenberg 2+ 2+ 0.28S. zuerich 2+ 2+ 0.22Citrobacter freundii 0.24 0.18 0.09Enterobacterfreundii X539A 0.25 0.20 0.03Enterobacter freundii 0.15 0.24 0.15Escherichia coli 2B14 0.34 0.34 0.18

a Control wells received no Spicer-Edwards IgG,but did receive goat anti-rabbit-alkaline phosphataseconjugate.

ed in microtitration plates with IF, ES, and thePCT was made by using 98 food samples. In thisstudy, the preenrichment medium was also ex-amined by EIA by using the following proce-dure: 0.1 ml of each 18- to 24-h preenrichmentculture was transferred to 10 ml of M-broth. TheM-broth subcultures were incubated for 4 to 6 hat 35°C and then were analyzed by EIA. Table 3contains data from individual subsamples ofnaturally contaminated foods that resulted inpositive identification of salmonellae by EIA.These foods all contained less than one viableSalmonella cell per gram. Thirteen subsampleswere positive by all methods. The Salmonellaserotypes from the chicken, yeast, and spicesamples were S. heidelberg, S. minnesota, andS. rubislaw, respectively. Five other subsamplesof these same foods were positive by one ormore IF test, but were negative by all othermethods; these probably were false-positive re-actions. False-positive IF reactions were alsoobtained on subsamples of pork liver and choco-late candy, but no salmonellae were detected byany other method. Because of the tendency ofIF to give false-positive reactions, we examinedclosely those instances when EIA yielded posi-tive results in the absence of conclusive culturalevidence for the presence of salmonellae (sam-ples 2, 4, 7, 9, 10, 11, and 19, Table 3).Two criteria were used to score positive EIA

samples. During the analysis of a large numberof pure cultures and food samples in microtitra-tion plates, we noted that control values varied,with respect to absorbance, from 0.06 to 0.38.

TABLE 2. Absorbancies obtained by using a FlowTitertek instrument when pure cultures were

analyzed by an EIAAbsorbance

OrganismSample Controla

Salmonella anatum 1.01 0.03S. rubislaw 1.27 0.03S. senftenberg 1.34 0.07S. zuerich 1.25 0.03Citrobacterfreundii 0.26 0.01Escherichia coli B 0.16 0.02

a Control wells received no Spicer-Edwards IgG,but did receive goat anti-rabbit-alkaline phosphataseconjugate.

Hence, an absorbance above 0.40 was one crite-rion that was used to conclude that a sample waspositive. With a few food samples, however, thedifference between the control and replicate testwells was appreciable, although the samplesgave absorbance values of <0.40. Therefore, adifference of 0.20 absorbance unit between con-trol and sample wells was considered a border-line positive. If these interpretations were cor-rect, then subsamples 2 and 4 of Table 3contained S. heidelberg detected only by EIAand, in one instance, only by EIA of the preen-richment broth. Likewise, subsamples 7, 9, 10,11, and 19 may have contained low levels of S.minnesota that did not survive selective enrich-ment. An alternative explanation is that thesesubsamples did not contain salmonellae and thatfalse-positive results were obtained by EIA.These possible false-positive EIA results couldhave been eliminated if the sole criterion of apositive test was an absorbance reading of 0.40or higher. But we did not choose to neglect theseborderline cases because they may be true posi-tive tests, and they represented 25% of thesubsamples. Also, these borderline reactionswere never observed when 100 or more subsam-ples of food that did not contain Salmonellawere examined. In all likelihood, each labora-tory will have to set its own parameters, basedon the sensitivity of the instrument, propertiesof the reagents, and type of food involved.

Sensitivity thresholds of EIA. Results obtainedby using three Salmonella serotypes for sensitiv-ity threshold determinations indicated that ap-proximately 106 Salmonella cells per ml (105cells per microtitration plate well) gave a strong-ly positive reaction that was sufficient for visualdiscrimination. S. rubislaw gave a weakly posi-tive reaction at the level of 105 cells per ml (104cells per microtitration plate well). Thus, thelimit of detection of salmonellae by EIA was asgood as, if not better than, the lower limit ofsensitivity of IF.A single experiment was conducted by using a

APPL. ENVIRON. MICROBIOL.

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

EIA FOR FOOD-BORNE SALMONELLAE 881

FIG. 3. Microcuvette showing representative Sal-monella-positive and -negative samples. Samples 1(wells 1 to 3) and 2 (wells 4 to 6) are positive tests asindicated by the chromogenic reaction product. Sam-ple 3 (wells 7 to 9) is negative. Wells 3, 6, and 9 arecontrol wells that received goat anti-rabbit-alkalinephosphatase conjugate but no Spicer-Edwards IgG.

fluorogenic substrate, 4-methyl umbelliferylphosphate. When S. anatum was used as the testorganism, 105 cells per ml were sufficient for apositive identification. Also, fluorogenic reac-tions were easier to read than the yellow reac-tion product of p-nitrophenyl phosphate, and noinstrumentation was necessary to detect a posi-tive reaction.

DISCUSSIONEnzyme-labeled antibodies are powerful as-

say tools (2, 8, 12, 19), and two publicationshave appeared on the detection of salmonellaewith EIA (5, 14). Krysinski and Heimsch (5)advocated the use of goat anti-rabbit immuno-globulin-horseradish peroxidase conjugates inan indirect test to detect salmonellae fixed tomembrane filters. Swaminathan and Ayres (14)developed a direct EIA for detecting salmonel-lae in foods; cell preparations were fixed to glassslides and were observed microscopically, as inIF. Although their method is excellent, the mi-croscopy requirement is a deterrent to wide-spread adoption of the method.Our original approach was to improve the

method of Krysinski and Heimsch (5) and to usecellulose acetate membrane filters and peroxi-dase conjugates. First, we purified IgG fromcommercial polyvalent Spicer-Edwards antiserato obtain a reagent free of cross-reactive anti-bodies (Fig. 1 and 2). This approach was basedon reports that 0 antigens stimulated almostentirely an IgM response in rabbits (6, 7) andflagellar antigens elicited primarily an IgG re-sponse (1, 10). We expected that the use ofpurified H-antibodies would solve what seemedto be a problem caused by cross-reactions be-tween nonsalmonellae and antibodies to somaticantigens. Later, we discovered that most of theapparent nonspecificity encountered was causedby the expression of intracellular peroxidases orcatalases or both in a variety of food-bornebacteria. This latter problem could not be over-come, even though several methods were usedto inactivate the bacterial enzymes before EIA

FIG. 4. Microtitration plate showing representa-tive Salmonella-positive and -negative samples. Sam-ples (three wells per sample) were added to verticalcolumns 1, 3, 5, 7, 9, and 11. Wells in rows C and Hthat received test material were used as controls.Wells in rows D and E were not used. Of the 12samples tested here, 3 were strongly positive (verticalcolumns 3, 7, and 11).

analysis. For example, fixed cells were treatedwith NaN3, 30% H202 in absolute methanol,formaldehyde vapors, or by heating. None ofthese treatments was successful. Furthermore,cell mats frequently sloughed off the membranesduring manipulation, which could lead to false-negative results.

Consequently, we used alkaline phosphataseconjugates. Since the products of alkaline phos-phatases on various substrates are soluble, un-like the product of peroxidase, the productswould not remain localized on membrane filters.Therefore, microtitration plate assays were de-veloped. Fixation of cells to the polystyrene ofmicrotitration plates likewise posed a problem.Although success was obtained in preventing thesloughing of cells in pure culture studies (S. A.Minnich, Ph.D. thesis, Iowa State University,Ames, 1980), inconsistencies were encounteredwhen foods were examined. This problem wasonly partly solved by using a direct EIA proce-dure (4) to reduce the number of wash stepsrequired. Consequently, we adapted the methodof Polin and Kennett (12), which does not re-quire cell fixation because unbound antibodyand conjugated antibody were removed by washsteps that used low-speed, short-time centrifuga-tion.The results obtained in pure culture studies

with the indirect EIA procedure indicated thatthe procedure was sensitive and specific. Eithervisual examination or absorbance determina-tions could be used to differentiate positive fromnegative samples (Tables 1 and 2).

Naturally contaminated foods that containedvery low numbers of salmonellae were thenanalyzed to compare the modified EIA with thethree most popular Salmonella screening proce-

VOL. 43, 1982

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

882 MINNICH, HARTMAN, AND HEIMSCH

TABLE 3. Subsamples of naturally contaminated foods that were positive by EIA and results obtained byusing IF, ES, and a PCT

Subsample EIA (absorbance) IFa ES PCTFood no. Preb Tet' SCd Tet SC Tet SC Tet SC

Chicken giblets 1 1.24 0.61 0.39 4+ 3+ + + + +2 0.82 0.16 0.13 - - - - - -3 0.50 0.19 0.59 3+ 3+ + + + +4 0.59 0.51 0.23 - 2+ - - - -5 0.45 0.14 0.41 - 4+ - + + +6 0.46 0.35 0.39 4+ 4+ + + + +

Dried yeast 7 0.35 0.16 ()e - -

8 1.26 0.61 (+) 4+ 4+ + + + +9 0.26 0.04 (-) 2+ - - - - -10 0.33 0.06 (-) - - - - - -11 0.34 0.07 (-) - - - - - -

Spice condiment 12 0.77 0.83 0.56 4+ 4+ + + + +13 0.91 1.45 0.79 4+ 4+ + + + +14 0.90 0.90 1.02 4+ 4+ + + + +15 0.67 0.93 0.85 4+ 4+ + + + +16 0.73 1.05 1.26 4+ 4+ + + + +17 0.91 0.55 0.99 4+ 4+ + + + +18 0.54 0.62 0.83 4+ 4+ + + + +19 0.35 0.18 0.20 3+ - + - - -20 0.87 1.09 0.97 4+ 4+ + + + +

aIF results were scored: -, very weak; +, weak; 2+, doubtful positive; 3+, definitely positive; and 4+, verystrong. Only 3+ and 4+ reactions were considered positive in this study.

b Absorbance readings were obtained by EIA after overnight preenrichment (Pre).c Tet, Tetrathionate broth.d SC, Selenite-cystine broth.eNot determined quantitatively (+ or - by visual observation).

dures. The most striking discovery of this seriesof experiments was that the EIA procedurecould detect the presence of salmonellae inpreenrichment broths. All positive confirmedsamples were initially identified as positive byEIA when preenrichment broths were subcul-tured for 4 to 6 h in M-broth before the EIA test(Table 3). Furthermore, in many instances high-er absorbance values were obtained from thepreenrichment broths than from subcultures ofthe selective enrichment broths. Also, some"doubtful positive" preenrichment broths weredetected by EIA; these invariably were obtainedfrom foods, subsamples of which had yieldedsalmonellae upon other occasions. Doubtfulpositive EIA tests were never obtained on sub-samples of foods, other subsamples of which didnot yield a Salmonella sp. by culture. Theseresults suggest that selective enrichment may betoo selective against many salmonellae and thatpreenrichment broths should be examined forthe presence of salmonellae. The EIA test issufficiently sensitive to do this. The fluorescent-antibody and possibly the ES tests also maypossess sufficient sensitivity, but the fluores-cent-antibody test would yield too many false-positive reactions and no one has studied theefficacy of ES tests when applied to preenrich-ment broths.

Both instruments used to measure productcolor (the Gilford and Flow Laboratories units)were satisfactory. Gilford microcuvettes hold0.5 ml in deep wells; therefore, cells could bethoroughly resuspended by adding 0.3 ml ofbuffer and gently shaking the cuvettes. In con-trast, the microtitration plates that we used hada well capacity of 0.25 ml. Occasionally, the useof wooden applicator sticks was required toresuspend the cell pellets.We obtained consistent and reproducible EIA

results when the M-broth was used in all posten-richment procedures. Similarly, good EIA re-sults were obtained by another laboratory whenM-broth was used (F. Delacroix, personal com-munication), but abnormalities arose when brainheart infusion or Trypticase soy broth was usedfor postenrichment. Problems may also arisewith various brands of microtitration plates;therefore, we recommend standardization of theEIA procedure with pure cultures before foodsamples are tested.The EIA procedure described here has several

advantages over other Salmonella screeningmethods. Sensitivity of the EIA is equivalent toor better than the lower limit of IF of 105salmonellae per ml (16). Tube agglutination, asperformed in ES, requires an optimal concentra-tion of about 107 salmonellae per ml (13), and

APPL. ENVIRON. MICROBIOL.

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from

EIA FOR FOOD-BORNE SALMONELLAE 883

this may be a drawback when preenrichmentbroths are examined. Problems in IF analysesalso are encountered when high concentrationsof salmonellae are present (16) because de-creased fluorescence of cells is obtained in thepresence of high concentrations of antigen; thiswould not be a problem with EIA. Furtheradvantages of the EIA are that the results can bequantitated and the reaction product is stable. Incontrast, IF results are subjective, and the fluo-rescence fades rapidly when the slides are ex-posed to UV light. The use of purified IgG inEIA results in specific reactions, whereas poly-valent OH conjugates conventionally used in IFresult in a false-positive rate of about 9% (16).Lastly, EIA is amenable to automation for rapidmultisample analysis. Automated enzyme-linked immunosorbent assay systems are cur-rently in use, and the adaptation of the Salmo-nella EIA test to present commercial equipmentcould be accomplished readily.The present Salmonella EIA has two draw-

backs. A centrifuge is needed, and we proposedan indirect procedure that required two centrifu-gation steps. A direct test was shown to befeasible (4), but its routine use in the laboratoryis dependent upon reliable means of fixing cellsheets to microtitration plates. More potent anti-H antibody preparations also would be helpful.Lastly, the use of fluorogenic substrates, whichwas only briefly addressed in this study, showspromise of increased sensitivity. We are examin-ing some of these factors in further detail.

ACKNOWLEDGMENTS

This study was primarily supported by the Iowa Agricultureand Home Economics Experiment Station under Project 2377and, to a lesser extent, by the Idaho Agricultural ExperimentStation under Project 708-R670.

LITERATURE CITED

1. Ada, G. L., G. J. V. Nossal, J. Pye, and A. Abbot. 1964.Preparation and properties of flagellar antigens from Sal-monella adelaide. Aust. J. Exp. Biol. Med. Sci. 42:267-282.

2. Engvall, E., and A. J. Pesce (ed.). 1978. Quantitativeenzyme immunoassay. Scand. J. Immunol. 8(Suppl. 7):1-129.

3. Food and Drug Administration. 1978. Bacteriological ana-lytical manual for foods, 5th ed. Association of OfficialAnalytical Chemists, Washington, D.C.

4. Hartman, P. A., and S. A. Minnich. 1981. Automation for

rapid detection of salmonellae in foods. J. Food Prot.44:385-393.

5. Krysinski, E. P., and R. C. Heimsch. 1977. Use of en-zyme-labeled antibodies to detect Salmonella in foods.Appl. Environ. Microbiol. 33:947-954.

6. Landy, M., R. P. Sanderson, and A. L. Jackson. 1965.Humoral and cellular aspects of the immune response tothe somatic antigen of Salmonella enteritidis. J. Exp.Med. 122:483-504.

7. Landy, M., and W. P. Weidanz. 1964. Natural antibodiesagainst gram-negative bacteria, p. 275-290. In M. Landyand W. Braun (ed.), Bacterial endotoxins. Institute ofMicrobiology, Rutgers State University, New Brunswick,N.J.

8. Maggio, E. T. (ed.). 1980. Enzyme immunoassay. CRCPress, Inc., Boca Raton, Fla.

9. March, S. C., I. Parikh, and P. Cuatrecasas. 1974. Asimplified method for cyanogen bromide activation ofagarose for affinity chromatography. Anal. Biochem.60:149-152.

10. Nossal, G. J. V., G. L. Ada, and C. M. Austin. 1964.Immunogenic properties of flagella, polymerized flagellinand flagellin in the primary response. Aust. J. Exp. Biol.Med. Sci. 42:283-294.

11. Poelma, P. L., and J. H. Silliker. 1976. Salmonella, p.301-328. In M. L. Speck (ed.), Compendium of methodsfor the microbiological examination of foods. AmericanPublic Health Association, Inc., Washington, D.C.

12. Polin, R. A., and R. Kennett. 1980. Use of monoclonalantibodies in an enzyme immunoassay for rapid identifica-tion of group B Streptococcus types II and III. J. Clin.Microbiol. 11:332-336.

13. Sperber, W. H., and R. H. Deibel. 1969. Acceleratedprocedure for Salmonella detection in dried foods andfeeds involving only broth culture and serological reac-tions. AppI. Microbiol. 17:533-539.

14. Swaminathan, B., and J. C. Ayres. 1980. A direct immu-noenzyme method for the detection of salmonellae infoods. J. Food Sci. 45:352-355, 361.

15. Swandnathan, B., J. C. Ayres, and J. E. Williams. 1978.Control of nonspecific staining in the fluorescent antibodytechnique for the detection of salmonellae in foods. Appl.Environ. Microbiol. 35:911-919.

16. Thomason, B. M. 1976. Fluorescent antibody detection ofsalmonellae, p. 329-343. In M. L. Speck (ed.), Compendi-um of methods for the microbiological examination offoods. American Public Health Association, Inc., Wash-ington, D.C.

17. Thomason, B. M. 1981. Current status of immunofluores-cent methodology for salmonellae. J. Food Prot. 44:381-384.

18. Voller, A., D. Bidwell, and A. Bartlett. 1976. Microplateenzyme immunoassays for the immunodiagnosis of viralinfections, p. 506-512. In N. R. Rose and H. Friedman(ed.), Manual of clinical immunology, 1st ed. AmericanSociety for Microbiology, Washington, D.C.

19. Voller, A., D. Bidwell, and A. Bartlett. 1979. The enzyme-linked immunosorbent assay (ELISA). Nuffield Labora-tories of Comparative Medicine, The Zoological Societyof London, London.

20. Yolken, R. H., and P. J. Stopa. 1979. Enzyme-linkedfluorescence assay: ultrasensitive solid-phase assay fordetection of human rotavirus. J. Clin. Microbiol. 10:317-321.

VOL. 43, 1982

on March 7, 2021 by guest

http://aem.asm

.org/D

ownloaded from